Analyzing the Effectiveness of Face Detection Algorithms in Improving Face Recognition Accuracy for Student Attendance Recording in Project-Based Learning Models Darmanto and Eka Wahyudi
Politeknik Negeri Ketapang, Ketapang, Kalimantan Barat, Indonesia 78813
Abstract
This study aims to compare the performance of two face detection algorithms, MTCNN and RetinaFace, in the context of a face recognition-based attendance system. The evaluation was conducted to measure the effectiveness of both models in terms of detection speed and accuracy using a diverse image dataset. This dataset includes images with varying numbers of faces, lighting conditions, and capture angles. The research process involved preprocessing images for consistency, setting parameters according to model documentation, and testing both algorithms on the same dataset. The results indicate that MTCNN has an average detection time of 1.97 seconds, making it more suitable for applications requiring quick responses, such as real-time attendance systems. Conversely, RetinaFace took an average of 8.75 seconds but showed advantages in detecting a higher number of faces, especially in more complex or less ideal images. Under optimal conditions, both algorithms were able to detect the same number of faces. However, in images with low lighting or challenging angles, RetinaFace demonstrated better performance. The conclusion of this study is that MTCNN is more suitable for applications prioritizing speed, while RetinaFace is ideal for situations requiring high accuracy despite longer processing times. The choice of algorithm depends on the specific needs of the application and its operating environment.
Keywords: Face detection, MTCNN, RetinaFace, automatic attendance, face recognition.
